Different software, different results

[laureannegille]

Hello,

using the same database but different software, I get different results and was wondering where it can come from

My model is very simple : 2 levels, dependantvariable_ij = \beta0 + \beta1 x Metric + u_0j + e_ij

Here are the results I get :

Capture.PNG (32.73 KiB) Viewed 4197 times

As you can see, R and Matlab provide similar results contrary to MLwiN.
I have checked that the depending variable is well defined in all cases and that I use the same computation method (RIGLS in MLwin is the same as reml in Matlab according to the help documentation).

Do you think at any other source of getting such different results, even in their interpretation ?
Thank you for any advice you can provide

[ChrisCharlton]

We cannot see an obvious reason for this difference. Do you have an example worksheet that you would be able to share that exhibits this behaviour?